Analytica Chimica Ada, 179 (1986) 279-287

Elsevier Science Publishers B.V., Amsterdam - Printed in The Netherlands

SIMULTANEOUS MULTIWAVELENGTH DETECTION IN FLOW

INJECTION ANALYSIS

F. LAZARO, A RfOS, M. D. LUQUE DE CASTRO and M. VALCARCEL

Department of Analytical Chemistry, Faculty of Sciences, University of C6doba,
C&do ba (Spain)
(Received 17th July 1985)

SUMMARY

The potential of the use of the diode-array detector in conjunction with flow injection
analysis is outlined. Methods for multicomponent resolution can be based on the formation
of complexes absorbing at different wavelengths. The example given is the determination
of mixtures of copper(I1) and iron(II1) with a mixed l:lO-phenanthroline/neocuproine re-
agent. Amplification and dilution methods are based on the sum of the absorbances at
several wavelengths and on monitoring absorbances at wavelengths away from the absorp-
tion maximum, respectively. The determination of nitrite via the Griess reaction is used as
an example; the viable determination range is extended to 0.002-60.0 fig ml-’ nitrite.
Software for the implementation of the suggested methods is outlined.

Multiwavelength diode arrays have become quite widely used as detectors

in high-performance liquid chromatography (h.p.1.c.). The possibility of
monitoring the different components of an eluate at the characteristic wave-
length for each of them represents an important improvement in mixture
resolution [l-3], which has been exploited in the development of quantita-
tive methods in industrial [4], clinical [ 5, 61 and environmental [ 7, 81
chemistry. Several firms now provide these detectors along with sophisticated
software for their use in conjunction with h.p.1.c. Nevertheless, the possibili-
ties of a detection system of this type can be of more interest from a practi-
cal point of view when they are used directly without a prior separation
technique.
Flow injection analysis (f.i.a.) has been compared with h.p.1.c. [9, lo],
but it differs essentially in that there is no column to separate individual
components of a sample. Nevertheless, f.i.a. has various possibilities for
simultaneous determinations of several species [ll] . A logical step in the
development of f.i.a. is therefore its association with fast-scan detectors as a
means of augmenting its value in multicomponent determinations. This asso-
ciation has been implemented with use of a fast-scan electrochemical detec-
tor [ 121, but there seem to be few practical applications.
In the context of a general research program on simultaneous determina-
tions by f.i.a., the present paper deals with the use of a diode-array detector

0003-2670/86/$03.50 o 1986 Elsevier Science Publishers B.V.

280

in conjunction with f.i.a. The combination is shown to be valuable in appli-

cations of dilution and amplification methods, which allow determination
limits to be manipulated over very broad ranges.
Commercial software developed for use in h.p.1.c. is not suitable for the
purposes of f.i.a.; its sophistication complicates rather than facilitates
troubleshooting. Thus two programs were written for multicomponent reso-
lution as well as for dilution and amplification methods, which allow the fast
and convenient resolution of mixed samples.

EXPERIMENTAL

Apparatus
A Hewlett-Packard 8451A diode-array detector was used with the HP-9121
floppy-disk unit, HP-98155A keyboard and HP-7470A plotter. Other equip-
ment included a Gilson Minipuls-2 peristaltic pump, Hellma 178.12-QS flow
cell (inner volume 18 PI), Tecator L-100-1 injection valve and Tecator TM-II
chemifold.

Simultaneous determination of iron or iron(III) and copper(H)

The reagent solution is a mixture of l,lO-phenanthroline (0.1% w/v) and
neocuproine (0.15% w/v) in 4:1 water/ethanol. The buffer solution contains
25 g of sodium acetate and 25 g of hydroxylammonium chloride in 250 ml
of water. Standard solutions (1.00 g 1-l) of Fe(II1) and Cu(I1) were prepared.
The manifold is shown in Fig. l(a). The sample is injected into the buf-
fered reducing stream, which ensures the reduction of the metal ions to Fe(I1)
and Cu(1) along the coil L1. This stream then merges with the reagent stream,
producing the indicator reactions in coil Lz.
The wavelengths of maximum absorption are 454 nm and 512 nm for
copper and iron, respectively and the detector monitors these two wave-
lengths simultaneously.
The optimization of variables yielded the results given above for the che-
mical variables. The optimum values found for the variables in f.i.a. were:
q1 = 1.7 ml min-‘, qz = 1.1 ml min-‘, Ll = 40 cm, L2 = 110 cm (0.5 mm i.d.),
and V, = 156.8 ~1.
The basis for the calculation program used for the resolution of these mix-
tures (MIXFIA program) is the establishment of n equations in m unknowns
(n > m), where n is the number of wavelengths at which the reaction is moni-
tored and m is the number of components in the mixture. A flow diagram of
this program is shown in Fig. 2. It works by collecting the maximum absor-
bance of the flow peak at n wavelengths per unknown sample. Prior to this,
one calibration curve is run per n value for each component (overall, n X m
equations).

Nitrite determination
The reagent solutions contain 0.95 g of sulphanilamide dissolved in 10 ml
of concentrated hydrochloric acid and diluted to 250 ml with water, and
 281

SAMPLE

NaAdNHgOH

o-PHENANTHROLINE
+
NEOCUPROINE

(bl

SAMPLE ! ,

WATER

SULPHANILAMIOE DIODE
+HCl ARRAY

N-II-NAPHTHYLI
ETHYLENEDIAMINE _ !
.NaCI W

Fig. 1. Manifolds used: (a) for resolution of Fe(III)/Cu( II) mixtures, (b) for nitrite deter-
mination. (For q, V, and L values, see text.)

0.36 g of N-( 1-naphthyl)ethylenediamine dihydrochloride and 2.9 g of NaCl

in 250 ml of water. A 1.00 g 1-l nitrite standard solution was prepared.
The manifold is shown in Fig. l(b). The sample (aqueous nitrite solution)
is injected into an aqueous stream which merges with the sulphanilamide solu-
tion, yielding the diazonium salt in coil LZ; this forms the monitored azo dye
when mixed with the N-(l-naphthyl)ethylenediamine stream in coil La. The
optimisation of variables yielded the above data for the chemical variables
and the following for the variables in the manifold: q1 = q2 = q3 = 1.3 ml
min-‘, L1 = 20 cm, L2 = 120 cm, L3 = 250 cm (0.5 mm i.d.), V, = 137.3 ~1.

RESULTS AND DISCUSSION

Simultaneous determination
The resolution of the mixture of iron(II1) or iron(I1) and copper(I1) ions is
based on the use of mixed selective reagents; l,lO-phenanthroline and neo-
cuproine were chosen for Fe(II1) and Cu(II), respectively. The working con-
ditions are similar for both complexes, whereas their u.v.-visible absorption
spectra are different.
The calibration equations obtained at the maximum absorption wave-
lengths for the two complexes (454 and 512 nm) were: for iron(I1) or
282

TABLE 1

Definition of variables (MIXFIA program)

Variable Meaning

A0 (I, Y) Intercepts of the calibration graphs

Al (I, Y) Slopes of the calibration graphs
A2 (Y, Ml) Sum of the maximum absorbances of the peaks obtained for each standard
at each wavelength chosen
A3 (J, Y) Average of the absorbances for each standard and wavelength
A4 (Y) This stores the absorbances at the maximum at the different working wave-
lengths
-408 Symbol of each component m the mixture
B$ Different from “Y” when no more injections of the standard are necessary
C (I, J) Concentration of standard J for component I of the mixture
CP When equal to “S”, standard measurements are made; if equal to “M” mix-
ture measurements are made
L(Y) Wavelengths at which the absorbances at the maximum of the f.i.a. peak are
obtained
Ll, L2 Initial and final wavelengths of the useful measurement range
M Number of wavelengths corresponding to the desired absorbance at the
maximum of the f.i.a. peak
Ml Number of injections for each standard
M4@ Different from “Y” when no more samples are analyzed
p (1) Number of standards used
Tl, T2 Initial and final measurement time from the moment of injection
T3 Integration time
T4 Interval between measurements
X(I) Results of the equation systems
Yl, Y2 Minimum and maximum absorbance values taken as the end of the y-axis of
the f.i.a. peak plot

iron(II1) in the range 1.0-8.0 pg ml-l,

AGs4 = 0.0626[Fe3’] - 0.0305 (r* = 0.995)
A 512 = 0.0858[Fe3’] - 0.0326 (r* = 0.996)
and for copper( II) in the range 15.0-30 pg ml-‘,
A 454 = O.O267[Cu*+] - 0.0907 (r’ = 0.999)
As12 = O.O044[Cu*+] - 0.0306 (r’ = 0.955)

It can be seen that the sensitivity for the Cu(I1) determination is much
lower than that of the conventional method. This is due partly to the fact
that in the conventional method, extraction of the chelate into chloroform
is used to increase sensitivity [ 131 and partly to the fact that the presence of
l,lO-phenanthroline dramatically diminishes the absorbance of the Cu(I)/
neocuproine system (by about 55%), because it competes for the Cu(1).
In the resolution of the mixture, it was necessary to take into account
 283

Fig. 2. Flow diagram of the MIXFIA program used for resolution of Fe(III)/Cu(II) mix-
tures. Variables are defined in Table 1.

that the individual absorbances are not additive in the mixture; the overall
absorbance is slightly higher than the sum of the individual values (8-10%).
This synergic effect required mathematical study and clarification. For this
purpose, calibration graphs were constructed for copper(I1) in the presence
of different iron(II1) concentrations. The slopes and intercepts of these
straight lines were a linear function of the iron(II1) concentration. Finally
expressions were established similar to those found in other cases of synergy
[14, 151, but this time at the two different wavelengths:
A 454 = -0.0025[Fe3’] [Cu”‘] + 0.1307[Fe3’] + 0.03407[Cuz’] - 0.2776
A 512 = -0.0033[Fe3’] [Cu”‘] + 0.1628[Fe3’] + 0.01493[Cuz’] - 0.2836
284

From these equations, it is possible to solve synthetic iron(III)/copper(II)

mixtures satisfactorily, as shown in Table 2. The r.s.d. values for 11 samples
containing 5 pg ml-’ Fe(II1) and 19 rugml’ Cu(I1) were +1.69% and +1.07%,
respectively.

Dilution and amplification methods

The Griess reaction (Shinn’s modification [ 161) was chosen for the devel-
opment of these methods. The calculation program (DINAFIA) used for the

TABLE 2

Resolution of synthetic mixtures of iron(II1) and copper(I1) by the suggested method

Added (pg ml-‘) Found (kg ml-‘) Errors (%)

Fe( III) Cu( II) Fe( III) Cu( II) Fe(II1) Cu( II)

3.00 16.00 3.08 16.10 2.7 0.6

5.00 16.00 5.00 16.03 0.0 0.2
7.00 22.00 7.13 21.79 1.8 -0.9
3.00 19.00 2.86 18.85 -4.6 -0.8
3.00 25 00 3.05 25.04 1.6 0.1
5.00 25.00 5.00 25.06 0.0 0.2
1.00 19.00 1.01 19.12 1.0 0.6

TABLE 3

Definition of variables (DINAFIA program)

Variable Meaning

A0 (1) Intercept of calibration graph I

Al (1) Slope of calibration graph I
A3 (I) Average absorbance for the injection made
A4 (Y) Values of the maximum absorbance at the wavelengths chosen
A$ Analyte formula
BS “Y” if another injection is made
C (J) Concentration of standard J
C$ “Y” when the standards are injected, “M” when the sample is injected
Ll, L2 Initial and final wavelengths of the interval used
L (Y, 1) Wavelengths corresponding to the desired maximum absorbance for calibration
graph I
Counter of the number of injections
&) Number of wavelengths corresponding to the desired maximum absorbance
N$ “Y” if another sample is injected
p (1) Number of standards injected for the construction of calibration graph I
R2 (1) Correlation coefficient of the calibration graph I
Tl, T2 Initial and final measurement times taken from the moment of injection
T3 Integration time
T4 Interval between measurements
Yl, Y2 Minimum and maximum absorbances used for plotting the f.i.a. peak
 285

automated nitrite determination over a very wide concentration range (2.0-

60 000 ng ml-‘) is based on measurement of the absorbances provided by the
analyte at different wavelengths depending on its concentration. If the nitrite
concentration is high (>4.0 pg ml-‘), medium (0.2-4.0 pg ml-‘) or low
(<0.2 pg ml-‘), the reaction product is monitored at a wavelength distant
from the absorption maximum (A,,,), at the absorption maximum (A,,,) or
at a series of wavelengths around the absorption maximum (Z A), respectively.
In the third case, the absorbances measured are added.
The program (Fig. 3) starts by asking for the working conditions (integra-
tion time, interval between measurements, initial and final time, hdii, A,,,,

Fig. 3. Flow diagram of the DINAFIA program for normal, dilution and amplification
methods. Variables are defined in Table 3.
oh), then it requests the injection into the flow-injection system of each of
the standards for the three calibration graphs (dilution, normal and amplifi-
cation method in the above-mentioned ranges). Afterwards, the features of
each calibration graph are calculated and, finally, the program evaluates the
nitrite concentration in the unknown samples, using the appropriate calibra-
tion graph in each case.
The calibration graphs used for nitrite determination were obtained by
monitoring the reaction product at 450 nm (dilution method), 542 nm (nor-
mal method) and the set of Z:h15 = 528, 530, . , .554, 556 nm (amplification
method). Table 4 lists the characteristic of these graphs and several others
for the dilution and amplification methods. The following conclusions can
be drawn. The range of determinations is extended from 0.2-4.0 pg ml-’ for
the conventional method to 0.002-60.0 pg ml-’ for the method suggested
here. Although the possible range of determination is considerably shortened
as a result of using the amplification plots instead of the dilution plots, the
linearity ratio (quotient between the upper and lower determination limits)
is very similar in every case. The sensitivity of the procedure, as well as the
manipulation factor (ratio between the slopes of the dilution or amplifica-
tion calibration graphs and the normal graphs) in the dilution calibration
increases with the proximity of hdll to h,,,, whereas in the amplification
calibration it increases with the number of wavelengths considered.
The relative errors found on injecting 15 samples with concentrations ly-
ing over a very wide range into the flow-injection system, are listed in
Table 5. Such errors are small for all the samples, but even smaller in the
normal method (average relative error +0.28%) than in the dilution method
(?1.58%) or the amplification method (?1.50%). The precision of the
method was calculated separately from 11 different samples for each curve.
The relative standard deviations obtained and the concentrations used were

TABLE 4

Manipulation of sensitivity in the determination of nitrite

Method Determination Sensitivity Manipulation

tnm) rangea (rg ml-’ )’ factor
(pg ml-’ )

Dilution 450 2.0-60.0 0.019 0.10

470 0.8-20.0 0.048 0.20
510 0.4-8.0 0.172 0.75
Normal 542 0.2-4.0 0.226 1.00
Amplification x h,b 0.06-2.0 0.686 3.00
=hnC 0.004-0.2 3.504 15.00
= hUd 0.002+I.l 5.484 24.00

aThe regression coefficient was 0.999 in all cases. bZh, = 540, 542, 544 (nm). CCh,, =
528, 530,. .., 554, 556 (nm). d xkz5 = 518,520,. . . ,564,566 (nm).
 287

TABLE 5

Relative errors in the determination of nitrite over a wide concentration range

Added Found Error Added Found Error

(erg ml-’ ) (pg ml-l) (%) (Mg ml-’ ) (fig ml-’ ) (%)

0.0066 0.0067 1.5 2.000 2.005 0.2

0.020 0.0202 1.0 3.060 3.075 0.5
0.0673 0.0688 2.2 4.000 3.950 -1.2
0.080 0.078 -2.5 10.000 9.840 -1.6
0.100 0.1003 0.3 12.000 11.770 -1.9
0.600 0.597 -0.5 20.000 19.720 -1.4
0.800 0.801 0.1 30.600 31.150 1.8
1.200 1.201 0.1

as follows: dilution, +0.40% (25 pg ml-‘); normal method, *0.13% (2.0 c1g
ml-‘); amplification method, +O.lO% (0.15 pg ml-‘).

REFERENCES

1 B. Vandeginste, R. Essers, T. Bosman, J. Reijn and G. Kateman, Anal. Chem., 57

(1985) 971.
2 S. A. George and A. Maute, Chromatographia, 15 (1982) 419.
3 A. C. J. Drouen, H. A. H. Billiet and L. de GaIan, Anal. Chem., 57 (1985) 962.
4 K. W. Jost, T. Crispin and I. Halasz, Erdoel Kohle, Erdgas, Petrochem., 37 (1984) 178.
5 T. Takeuchi and D. Ishii, Chromatogr. Commun., 7 (1984) 151.
6 A. Fell, B. J. Clark and H. P. Scott, J. Chromatogr., 297 (1984) 203.
7 D. J. Desilets, P. T. Kissinger, F. E. Lytle, M. A. Horne, M. S. Ludwrczak and R. B.
Jacko, Environ Sci. Technol., 18 (1984) 386.
8 J. L. Anderson, K. K. Whiten, J. D. Brewster, T. Y. Ou and W. K. Nomdez, Anal.
Chem., 57 (1985) 1366.
9 J. RbZiEka and E. H. Hansen, Flow Injection Analysis, Wiley, New York, 1981
10 M. Vslckcel and M. D. Luque de Castro, Flow Injection Analysis: Principles and
Applications, Ellis Horwood, Chichester, in press.
11 M. D. Luque de Castro and M. VaicCcel, Analyst (London), 109 (1984) 413.
12 J. Janata and J. RRZiEka, Anal. Chim. Acta, 139 (1982) 105.
13 A. R. Gahler, Anal. Chem., 26 (1954) 577.
14 A. Rios and M. Vai&cel, Talanta, 32 (1985) 851.
15 A. Rios, M. Silva and M. VaicCcel, Z. Anal. Chem., 320 (1985) 762.
16 M. B. Shinn, Ind. Eng. Chem. Anal. Ed., 13 (1941) 33.